Specific Emitter Identification Using HOG Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional specific emitter identification methods are limited in accurately distinguishing between emitters due to inconsistent emission patterns, particularly for transmitters with variable signatures, and often require multiple recordings to reliably identify devices.
Innovation Solution
The method employs Histogram of Oriented Gradient (HOG) features extracted from the time-frequency turn-on transients of emitters, allowing for robust identification by comparing computed HOG features with stored templates, independent of transmitted messages, and capable of identifying emitters with single or multiple recordings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional SEI matching methods are used, then emitter identification can be performed, but the methods are limited in accuracy for emitters with variable signatures and often require multiple recordings
Solution Approach 1:
The patent transforms the emitter signal into a time-frequency representation (spectrogram) and extracts HOG features from this transformed domain. This parameter transformation from raw signal to feature space enables more accurate identification of emitters with variable signatures, resolving the contradiction between identification accuracy and the number of recordings required.
Solution Approach 2:
The patent creates a template representation of emitter characteristics through HOG features from training recordings. This template serves as a reference copy that can be compared against new emitter signals, enabling accurate identification with fewer recordings while maintaining high precision for variable signature emitters.
2Reliability
If HOG features are extracted from time-frequency turn-on transients, then robust emitter identification is achieved independent of transmitted messages, but the processing complexity increases
Solution Approach 1:
The patent segments the time-frequency spectrogram into multiple cells or regions, computing HOG features for each segment independently. This segmentation approach enables robust emitter identification by capturing local patterns while reducing the overall processing complexity through divide-and-conquer strategy.
Solution Approach 2:
The patent extracts only the essential gradient orientation information from the time-frequency representation, discarding redundant signal details. This extraction of critical features (HOG descriptors) maintains identification robustness while significantly reducing processing complexity compared to analyzing the complete raw signal.
Data Source
AI summary
In one embodiment, a method for specific emitter identification includes receiving a signal from an emitter indicative of a hardware characteristic of the emitter. A computer-readable representation of the received signal is generated. A plurality of gradients for each partition of a plurality of partitions of the computer-readable representation is computed. Each gradient is indicative of at least the angular orientation of a respective portion of the computer-readable representation. A histogram is computed for each partition by assigning each computed gradient to a bin based at least in part on the magnitude of the computed gradient. One or more Histogram of Oriented Gradient (HOG) features are extracted from a concatenation of the bins of all of the computed histograms. The one or more HOG features are compared to one or more corresponding HOG features stored on a computer-readable medium. Based at least in part on the comparison, a determination is made regarding whether the emitter has a particular identification.